[Fwd: Re: regarding libata HPA default behavior]

Tejun Heo htejun at gmail.com
Fri Sep 7 03:23:06 BST 2007


Luiz Fernando N. Capitulino wrote:
> Em Thu, 6 Sep 2007 14:38:48 +0100
> Matthew Garrett <mjg59 at srcf.ucam.org> escreveu:
> 
> | > Seems to me the dangers and damage caused by a suddenly-too-small disk 
> | > are greater than those caused by a suddenly-larger disk.  The latter 
> | > case only confuses schemes that are offset from the last reported LBA, 
> | > whereas the former can cause data loss and worse.
> | 
> | Indeed. When we transitioned to libata for pata devices, a large number 
> | of users suddenly had regressions. It's not just an issue for laptops, 
> | there's a moderate number of desktop drives that are clipped to 32GB in 
> | order to avoid breaking ancient Phoenix BIOSes. The traditional 
> | behaviour of drivers/ide has resulted in many people having file systems 
> | that extend into the HPA, and breaking those setups wasn't acceptable 
> | for us. As a result, we extend the disk by default in Ubuntu.
> 
>  Althought Mandriva still provides default libata settings, I'm
> with Matthew here and I'm considering to make that change.
> 
>  Tejun, I think that by trying to be compatible with everything
> will cause more trouble than fixes.
> 
>  We'll have to make a choice, I think the one that causes less
> troubles is to maintain compability with the IDE drivers.
> 

I think I'll make SUSE kernel to default to unlocking HPA too.  Ewwww...
This whole HPA thing got very ugly and I'm not still sure if this is the
right way to go.  Oh well, lesser of the two evils, I guess.

Thanks.

-- 
tejun



More information about the kernel-team mailing list